Abstract
The performance of multi-stage production and transportation systems depends heavily on the ordering system that determines the order release for each process. Various ordering systems have been developed, and the just-in-time (JIT) ordering system is a typical system. Under JIT ordering systems, the order release for each process is determined on the basis of actual demand, that is, without demand forecasts, and thus avoids the influence of forecasting error. However, the influence of demand changes cannot be avoided even in the JIT ordering systems. In the previous literature on JIT ordering systems, demand changes have been considered, and stable changes in product demand have been assumed in most of the previous literature on JIT ordering systems. Based on the background, reactive JIT ordering systems were developed, and it was shown that the systems developed are effective for responding to unstable changes in product demand. In the reactive JIT ordering systems, the buffer sizes for all of the stages are controlled by a central controller. One purpose of JIT ordering is the decentralization of control, and decentralized systems are more suitable for JIT production than centralized systems. Therefore, this paper proposes a decentralized control method for reactive Kanban system for responding to unstable changes in product demand. In the proposed method, unstable changes in the demand from the succeeding stage can be detected at every stage, and a response to the detected changes can be determined individually at every stage. A mathematical model of the multi-stage production and transportation system controlled by the Kanban system is formulated, and it is decomposed into single-stage processing systems controlled by the Kanban system. The performance of the decomposed systems is analyzed by queuing theory, and a control rule is developed on the basis of the analysis. The performance of the proposed method is investigated by simulation experiments, and the findings obtained in this paper are summarized as conclusions.
